Field notes. John Arndt, Soxoa. Published .
The AI Readiness Assessment is fifteen questions across five dimensions. It runs in your browser, stores nothing, and takes about five minutes. It exists because the question people open with is usually which tool should we buy, and that is almost never the question that decides how this goes.
Here is what it looks at, in its own words, and why the first dimension carries the other four.
The five dimensions
Each dimension is scored on its own, and the result names your weakest one with a first move attached. The dimensions and their first moves are printed below exactly as the assessment states them, because a diagnostic you cannot audit is not worth taking.
- Ownership
- “Whether a named person is accountable for AI decisions and has the authority to make them.” First move: “Name one accountable owner for AI decisions and give them a standing slot with leadership. Nothing else holds without this.”
- Workflows
- “Whether the work is understood well enough to know what is worth changing.” First move: “Pick one recurring workflow, write down its steps, volumes, and exceptions, and agree what a better result would look like.”
- Data & tools
- “Whether the information and approved tools exist for a system to do useful work safely.” First move: “Decide which tools are approved for which kinds of information, and make real examples of the work available to test against.”
- People
- “Whether the team can use, question, and maintain what gets built.” First move: “Get the people who do the work building with approved tools on their own tasks, with someone experienced in the room.”
- Governance
- “Whether review, risk, and measurement are decided before something goes wrong.” First move: “Write a one-page use policy, define who reviews AI output before it counts, and choose one measure you will actually check.”
Why ownership is the dependency
Read the other four first moves again and notice what they have in common. Pick one workflow. Decide which tools are approved for which information. Get people building. Write the policy and choose the measure. Every one of them is a decision, and a decision needs somebody with the authority to make it and the standing to make it stick.
That is why ownership is not one factor among five. It is the one the others rest on. An organization with strong workflow documentation, good data, capable people, and no owner produces exactly the pattern I see most often: several teams doing sensible things separately, no agreement about which tools may touch which information, and a policy document that describes a process nobody uses. Conversely an organization with a genuine owner and weak everything else makes progress, because the owner can get the workflow written down, the tool decision made, and the measure chosen. Those are all things a person with authority can do in a quarter.
Ownership is also the dimension people are most tempted to fake. A steering group is not an owner. Every member holds an effective veto and none of them owns the outcome, so the safe move is always to ask for more review, and the predictable result is a policy, a pilot that never ends, and a set of tools the staff use anyway without telling anybody. One name beats a committee, chiefly because one name can decline things and be held to the decision later.
In the scoring, ties resolve toward ownership for that reason: if ownership is level with something else at the bottom, ownership is the one named as weakest.
What the result actually tells you
The overall score maps to one of four stages, each with its characteristic risk rather than a grade. Exploring: interest is real but nothing is decided, and the risk is drifting into scattered tool use with no owner and no result. Piloting: you know where the work is and somebody cares, and the risk is a pilot that never becomes how the work is done. Operating: something runs in normal operations, and the risk is that it depends on one person while nobody is steering what comes next. Scaling: AI is part of how you operate, and the risk is sprawl, meaning more systems than oversight and measures that stopped being checked.
Then it recommends a next step and prints the rule it used, so you can disagree with it on the evidence. If something already runs in routine use and steering is what trails, it points at a fractional lead. If the work and the owner are both reasonably solid and nothing is in routine use yet, it points at a build sprint. Otherwise it points at a workshop, on the theory that a team which has done it once with someone experienced in the room argues about the right things afterwards.
When the honest result is buy nothing yet
One combination returns no engagement at all: nobody is accountable for AI decisions, nobody can name the recurring work that costs the most time, and the overall score is low. In that case the assessment says to buy nothing yet, and it means it. Nothing on the ladder is recommended from that sheet.
The reason is not modesty. Buying in that state produces the licence-and-a-shrug outcome, and it spends the organization’s appetite for trying this at all. The two prerequisites are cheap and neither of them needs a vendor: name one person who is accountable, and write down one recurring workflow with its steps, volumes, and exceptions. Both are a fortnight of ordinary work. Do them and take the assessment again; the answer will be different, and it will be different because the underlying situation is.
If your result does recommend something, treat it as an argument rather than a verdict. The rule is printed for exactly that reason, and the fifteen questions cannot know about the reorganisation you are three weeks away from announcing.
Where this fits
Take the AI Readiness Assessment: it is free, it runs entirely in your browser, and your answers are not stored or sent anywhere. If you want to argue with the result, or you already know ownership is the gap and want to talk about what to do about it, schedule a strategy call. Scope and price for any engagement are agreed before work begins.